Transcript Health Informatics

Health Recommender System
Undergraduates: Alex Rogers, Austin Buono, Kate Hawkins, & Tom Hortemiller
Mentors: Kelly Caine & Fuxiao Xin
Endnote
One of the most valuable learning experiences of our research was
becoming familiar with Endnote. Endnote is a commercial reference
management software package, used to manage bibliographies and
references when writing essays and articles.
Citing sources is one of the most valuable components of a research
paper. Giving people insight to where the information originated is an
invaluable step towards further development in the future.
Endnote was an extremely useful tool in our collaboration throughout
the project. We met several times to discuss research we had found
individually, and having our sources neatly recorded made it easy to
share information with each other.
IRB
Since we knew a project like ours would likely involve sensitive data
such as patient information at some point in time, we decided to look
into the institutional review board (IRB), also known as an independent
ethics committee or ethical review board. An IRB is a committee set up
to approve, monitor, and review any research that involves humans.
They aim to maintain ethical treatment and protect the rights and
welfare of all research subjects. In the United States, the FDA along with
the Department of Health and Human Services empower IRBs to
approve, require modifications in planned research prior to approval, or
disapprove research. An IRB performs critical oversight functions for
research conducted on human subjects that are scientific and ethical.
Not all forms of research require IRB approval.
Below is a checklist that we reviewed when determining whether or not
we would need IRB approval:
Is it research?
If BOTH of the following are true, your activity involves research:
-The activity is a systematic investigation, including research
development, testing and evaluation.
-The activity is designed to develop or contribute to general knowledge.
Does it involve human subjects?
Does my activity involve deceased individuals?
If it does, your activity requires submission to the IUPUI/Clarian IRB.
Does my research involve the FDA?
Is it a student project involving risk to human subjects?
TEMPLATE DESIGN © 2008
www.PosterPresentations.com
Research
Our approach to this research project was highly individual, with a
major group component at the end to develop our final product.
In the beginning, we met as a group once per week to discuss topics
related to our project that we had researched, and to determine which
direction we wanted to go next. We would then separate for the week.
During this time, we would do individual research over the ideas that we
wanted to address at our next meeting. This allowed us to have multiple
resources and results when we got together to collaborate with our
research. We would then discuss what we considered to be the best
sources of information to base our findings on, and then repeat this
process.
Web Based Health Recommender System
“Our recommendation system was designed with the goal of providing
accurate, low cost medical recommendations in countries where health
care costs are prohibitively expensive, this system can provide a free
alternative…like a second opinion.”
1. A database of rules
2. A database of facts
3. An inference engine
Extraction
while(<FL3>) {
chomp;
#this is parsing one line of the file
#escaping paring the first line, which is the field
PERSON_ID (PRIMARY KEY)
name
YEAR_OF_BIRTH (age)
if (not /^\d/) {
GENDER_COCEPT_ID (gender)
next;
SOURCE_RACE_CODE (race)
}
DRUG_CONCEPT_ID (medication taken) #split the line into each field, and put the value of
CONDITION_CONCEPT_ID (side effect) each field into an array
my @line = split(/\t/, $_);
my $Condition_Concept_ID = $line[5];
$personID = $line[2];
We found this to be a very successful research method, and were then
able to apply what we had found to our project as a group. Reviewing
multiple resources gave us a base to build our ideas on, and allowed us
to be more knowledgeable when approaching our final task.
if (not exists $line[5]) {
next;
}
$data{$personID}{'Side_Effects'}{
$Condition_Concept_ID}=1 ;
All of this gave us a better understanding and respect for the
importance of research.
}
Recommender Systems
Provide advice to users about items they wish to purchase or examine.
The recommendations made by the system can help users search
through large amounts of information. Also, the system needs to be able
to make a connection between the user’s needs and the available items
that are in the system to be recommended. A key part to a recommender
system is the knowledge of the content that is in the system. The system
needs to be able to make a connection between the user’s needs and the
available items that are in the system to be recommended
There are 4 different recommender techniques that are widely used:
1. The collaborative recommendation system has knowledge through
collaborative opinion profiles, demographic profiles and user opinions
2. The content- based recommendation has the task of learning a
specific classification rule for each user on the basis of the user’s rating
information and the attributes of each item so that items can be
classified as likely to be interesting or not
3. The knowledge based- recommender system has an emphasis on the
user’s situation and how recommended items can meet that particular
need
4. A hybrid recommender is one that uses recommendation components
or logic of different types. For example, it may combine the methods of
collaborative and content-based recommender systems
Left Side-These rules create the fact database which supports the
Inference Engine; tasked with providing recommendations based on the
user’s input and the known facts. The fact database also contains the
information required by the explanation system, which helps users
understand their recommendation.
Right Side-A model of the entire system on a single machine
Carbonell, J., Siekmann, J., An, A., Stefanowski, J., Ramanna, S., Butz, C., et al. (2007). Web Based Health
Recommender System Using Rough Sets, Survival Analysis and Rule-Based Expert
Systems. (p. 491). doi:10.1007/978-3-540-72530-5&#95;59.
Data Set
Orwik 2009/2010 OMOP Cup: Methods Competition
-current methods for detecting adverse drug events from observational
data are undeveloped
-helps to ensure the safety of prescription drugs
-a major global public health challenge
-there is a need for new data sources and new algorithms to respond to
the challenge
-Each challenge offers a Grand Prize
S_person: contains a unique identifier for each person along with
demographic data.
S_ObsPeriod: contains the span of time for which data is captured in the
database for a given patient.
S_DrugExp: contains periods of drug usage for each person.
S_ConditionOCC: contains records of conditions and when they were
observed.
K-nearest Neighbor
After the extraction of the data, this information must be ran through
some sort of algorithm to distinguish what medications would be right
for a certain person. The algorithm that we chose was K Nearest
Neighbor and is the State-of- the- art algorithm used in recommender
systems. With this algorithm we are capable of finding similarities
between patients and predicting the best medication for them. Using
this algorithm also allowed use to create this type of vision with the data
we extracted:
Goals
-Become familiar with research methods
-Understand research process
-Gain experience in conducting research
-Learn to distinguish credibility of sources
-Explore the Health Informatics field
-Understand state of the art recommender systems
-Apply knowledge from research to real world data set
POSTER TEMPLATE BY:
www.PosterPresentations.com
Health Recommender System
Endnote
Undergraduates: Alex Rogers, Austin Buono, Kate Hawkins, & Tom Hortemiller
Mentors: Kelly Caine & Fuxiao Xin
Endnote
One of the most valuable learning experiences of our
Research
Web Based Health Recommender System
research was becoming familiar with Endnote. Endnote is a
commercial reference management software package, used
to manage bibliographies and references when writing
essays and articles.
One of the most valuable learning experiences of our research was
becoming familiar with Endnote. Endnote is a commercial reference
management software package, used to manage bibliographies and
references when writing essays and articles.
Citing sources is one of the most valuable components of a research
paper. Giving people insight to where the information originated is an
invaluable step towards further development in the future.
In the beginning, we met as a group once per week to discuss topics
related to our project that we had researched, and to determine which
direction we wanted to go next. We would then separate for the week.
During this time, we would do individual research over the ideas that we
wanted to address at our next meeting. This allowed us to have multiple
resources and results when we got together to collaborate with our
research. We would then discuss what we considered to be the best
sources of information to base our findings on, and then repeat this
process.
“Our recommendation system was designed with the goal of providing
accurate, low cost medical recommendations in countries where health
care costs are prohibitively expensive, this system can provide a free
alternative…like a second opinion.”
1. A database of rules
2. A database of facts
3. An inference engine
Citing sources is one of the most valuable components of a
research paper. Giving people insight to where the
information originated is an invaluable step towards further
Recommender Systems
development in the future.
Endnote was an extremely useful tool in our collaboration throughout
the project. We met several times to discuss research we had found
individually, and having our sources neatly recorded made it easy to
share information with each other.
IRB
Our approach to this research project was highly individual, with a
major group component at the end to develop our final product.
Since we knew a project like ours would likely involve sensitive data
such as patient information at some point in time, we decided to look
into the institutional review board (IRB), also known as an independent
ethics committee or ethical review board. An IRB is a committee set up
to approve, monitor, and review any research that involves humans.
They aim to maintain ethical treatment and protect the rights and
welfare of all research subjects. In the United States, the FDA along with
the Department of Health and Human Services empower IRBs to
approve, require modifications in planned research prior to approval, or
disapprove research. An IRB performs critical oversight functions for
research conducted on human subjects that are scientific and ethical.
Not all forms of research require IRB approval.
Below is a checklist that we reviewed when determining whether or not
we would need IRB approval:
Does it involve human subjects?
Does my activity involve deceased individuals?
If it does, your activity requires submission to the IUPUI/Clarian IRB.
Does my research involve the FDA?
Is it a student project involving risk to human subjects?
TEMPLATE DESIGN © 2008
www.PosterPresentations.com
while(<FL3>) {
chomp;
#this is parsing one line of the file
#escaping paring the first line, which is the field
PERSON_ID (PRIMARY KEY)
name
YEAR_OF_BIRTH (age)
if (not /^\d/) {
GENDER_COCEPT_ID (gender)
next;
SOURCE_RACE_CODE (race)
}
DRUG_CONCEPT_ID (medication taken) #split the line into each field, and put the value of
CONDITION_CONCEPT_ID (side effect) each field into an array
my @line = split(/\t/, $_);
my $Condition_Concept_ID = $line[5];
$personID = $line[2];
We found this to be a very successful research method, and were then
able to apply what we had found to our project as a group. Reviewing
multiple resources gave us a base to build our ideas on, and allowed us
to be more knowledgeable when approaching our final task.
All of this gave us a better understanding and respect for the
importance of research.
Provide advice to users about items they wish to purchase or examine.
The recommendations made by the system can help users search
through large amounts of information. Also, the system needs to be able
to make a connection between the user’s needs and the available items
that are in the system to be recommended. A key part to a recommender
system is the knowledge of the content that is in the system. The system
needs to be able to make a connection between the user’s needs and the
available items that are in the system to be recommended
There are 4 different recommender techniques that are widely used:
1. The collaborative recommendation system has knowledge through
collaborative opinion profiles, demographic profiles and user opinions
2. The content- based recommendation has the task of learning a
specific classification rule for each user on the basis of the user’s rating
information and the attributes of each item so that items can be
classified as likely to be interesting or not
3. The knowledge based- recommender system has an emphasis on the
user’s situation and how recommended items can meet that particular
need
4. A hybrid recommender is one that uses recommendation components
or logic of different types. For example, it may combine the methods of
collaborative and content-based recommender systems
if (not exists $line[5]) {
next;
}
$data{$personID}{'Side_Effects'}{
$Condition_Concept_ID}=1 ;
}
Left Side-These rules create the fact database which supports the
Inference Engine; tasked with providing recommendations based on the
user’s input and the known facts. The fact database also contains the
information required by the explanation system, which helps users
understand their recommendation.
Right Side-A model of the entire system on a single machine
Endnote was an extremely useful tool in our collaboration
throughout the project. We met several times to discuss
Data Set
research we had found individually, and having our sources
neatly recorded made it easy to share information with
each other.
Is it research?
If BOTH of the following are true, your activity involves research:
-The activity is a systematic investigation, including research
development, testing and evaluation.
-The activity is designed to develop or contribute to general knowledge.
Extraction
Carbonell, J., Siekmann, J., An, A., Stefanowski, J., Ramanna, S., Butz, C., et al. (2007). Web Based Health
Recommender System Using Rough Sets, Survival Analysis and Rule-Based Expert
Systems. (p. 491). doi:10.1007/978-3-540-72530-5&#95;59.
Orwik 2009/2010 OMOP Cup: Methods Competition
-current methods for detecting adverse drug events from observational
data are undeveloped
-helps to ensure the safety of prescription drugs
-a major global public health challenge
-there is a need for new data sources and new algorithms to respond to
the challenge
-Each challenge offers a Grand Prize
S_person: contains a unique identifier for each person along with
demographic data.
S_ObsPeriod: contains the span of time for which data is captured in the
database for a given patient.
S_DrugExp: contains periods of drug usage for each person.
S_ConditionOCC: contains records of conditions and when they were
observed.
K-nearest Neighbor
After the extraction of the data, this information must be ran through
some sort of algorithm to distinguish what medications would be right
for a certain person. The algorithm that we chose was K Nearest
Neighbor and is the State-of- the- art algorithm used in recommender
systems. With this algorithm we are capable of finding similarities
between patients and predicting the best medication for them. Using
this algorithm also allowed use to create this type of vision with the data
we extracted:
Health Recommender
System
IRB
Undergraduates: Alex Rogers, Austin Buono, Kate Hawkins, & Tom Hortemiller
Since we knew a project like ours would likely involve
dataXin
such as patient information at some
Mentors: sensitive
Kelly Caine & Fuxiao
point in time, we decided to look into the institutional review board (IRB), also known as an independent
Endnote
Research
Web Based Health Recommender System
Extraction
ethics committee or ethical review board. An IRB is a committee set up to approve, monitor, and review
any research that involves humans. They aim to maintain ethical treatment and protect the rights and
welfare of all research subjects. In the United States, the FDA along with the Department of Health and
Human Services empower IRBs to approve, require modifications in planned research prior to approval, or
disapprove research. An IRB performs critical oversight functions for research conducted on human
subjects that are scientific and ethical. Not all forms of research require IRB approval.
Is it research?IRB
If BOTH of the following are true, your activity involves research:
Recommender Systems
-The activity is a systematic investigation, including research development, testing and evaluation.
-The activity is designed to develop or contribute to general knowledge.
Does it involve human subjects?
One of the most valuable learning experiences of our research was
becoming familiar with Endnote. Endnote is a commercial reference
management software package, used to manage bibliographies and
references when writing essays and articles.
Citing sources is one of the most valuable components of a research
paper. Giving people insight to where the information originated is an
invaluable step towards further development in the future.
Endnote was an extremely useful tool in our collaboration throughout
the project. We met several times to discuss research we had found
individually, and having our sources neatly recorded made it easy to
share information with each other.
Since we knew a project like ours would likely involve sensitive data
such as patient information at some point in time, we decided to look
into the institutional review board (IRB), also known as an independent
ethics committee or ethical review board. An IRB is a committee set up
to approve, monitor, and review any research that involves humans.
They aim to maintain ethical treatment and protect the rights and
welfare of all research subjects. In the United States, the FDA along with
the Department of Health and Human Services empower IRBs to
approve, require modifications in planned research prior to approval, or
disapprove research. An IRB performs critical oversight functions for
research conducted on human subjects that are scientific and ethical.
Not all forms of research require IRB approval.
Below is a checklist that we reviewed when determining whether or not
we would need IRB approval:
Our approach to this research project was highly individual, with a
major group component at the end to develop our final product.
In the beginning, we met as a group once per week to discuss topics
related to our project that we had researched, and to determine which
direction we wanted to go next. We would then separate for the week.
During this time, we would do individual research over the ideas that we
wanted to address at our next meeting. This allowed us to have multiple
resources and results when we got together to collaborate with our
research. We would then discuss what we considered to be the best
sources of information to base our findings on, and then repeat this
process.
“Our recommendation system was designed with the goal of providing
accurate, low cost medical recommendations in countries where health
care costs are prohibitively expensive, this system can provide a free
alternative…like a second opinion.”
1. A database of rules
2. A database of facts
3. An inference engine
We found this to be a very successful research method, and were then
able to apply what we had found to our project as a group. Reviewing
multiple resources gave us a base to build our ideas on, and allowed us
to be more knowledgeable when approaching our final task.
if (not exists $line[5]) {
next;
}
$data{$personID}{'Side_Effects'}{
$Condition_Concept_ID}=1 ;
All of this gave us a better understanding and respect for the
importance of research.
Provide advice to users about items they wish to purchase or examine.
The recommendations made by the system can help users search
through large amounts of information. Also, the system needs to be able
to make a connection between the user’s needs and the available items
that are in the system to be recommended. A key part to a recommender
system is the knowledge of the content that is in the system. The system
needs to be able to make a connection between the user’s needs and the
available items that are in the system to be recommended
There are 4 different recommender techniques that are widely used:
1. The collaborative recommendation system has knowledge through
collaborative opinion profiles, demographic profiles and user opinions
}
Left Side-These rules create the fact database which supports the
Inference Engine; tasked with providing recommendations based on the
user’s input and the known facts. The fact database also contains the
information required by the explanation system, which helps users
understand their recommendation.
Right Side-A model of the entire system on a single machine
Carbonell, J., Siekmann, J., An, A., Stefanowski, J., Ramanna, S., Butz, C., et al. (2007). Web Based Health
Recommender System Using Rough Sets, Survival Analysis and Rule-Based Expert
Systems. (p. 491). doi:10.1007/978-3-540-72530-5&#95;59.
Does my activity involve deceased individuals?
If it does, your activity requires submission to the IUPUI/Clarian IRB.
Is it research?
If BOTH of the following are true, your activity involves research:
-The activity is a systematic investigation, including research
development, testing and evaluation.
-The activity is designed to develop or contribute to general knowledge.
2. The content- based recommendation has the task of learning a
specific classification rule for each user on the basis of the user’s rating
information and the attributes of each item so that items can be
classified as likely to be interesting or not
Does my research involve the FDA?
Does it involve human subjects?
Does my activity involve deceased individuals?
If it does, your activity requires submission to the IUPUI/Clarian IRB.
3. The knowledge based- recommender system has an emphasis on the
user’s situation and how recommended items can meet that particular
need
4. A hybrid recommender is one that uses recommendation components
or logic of different types. For example, it may combine the methods of
collaborative and content-based recommender systems
Is it a student project involving risk to human subjects?
Does my research involve the FDA?
Is it a student project involving risk to human subjects?
TEMPLATE DESIGN © 2008
www.PosterPresentations.com
while(<FL3>) {
chomp;
#this is parsing one line of the file
#escaping paring the first line, which is the field
PERSON_ID (PRIMARY KEY)
name
YEAR_OF_BIRTH (age)
if (not /^\d/) {
GENDER_COCEPT_ID (gender)
next;
SOURCE_RACE_CODE (race)
}
DRUG_CONCEPT_ID (medication taken) #split the line into each field, and put the value of
CONDITION_CONCEPT_ID (side effect) each field into an array
my @line = split(/\t/, $_);
my $Condition_Concept_ID = $line[5];
$personID = $line[2];
Data Set
Orwik 2009/2010 OMOP Cup: Methods Competition
-current methods for detecting adverse drug events from observational
data are undeveloped
-helps to ensure the safety of prescription drugs
-a major global public health challenge
-there is a need for new data sources and new algorithms to respond to
the challenge
-Each challenge offers a Grand Prize
S_person: contains a unique identifier for each person along with
demographic data.
S_ObsPeriod: contains the span of time for which data is captured in the
database for a given patient.
S_DrugExp: contains periods of drug usage for each person.
S_ConditionOCC: contains records of conditions and when they were
observed.
K-nearest Neighbor
After the extraction of the data, this information must be ran through
some sort of algorithm to distinguish what medications would be right
for a certain person. The algorithm that we chose was K Nearest
Neighbor and is the State-of- the- art algorithm used in recommender
systems. With this algorithm we are capable of finding similarities
between patients and predicting the best medication for them. Using
this algorithm also allowed use to create this type of vision with the data
we extracted:
Health Recommender
System
Research
Undergraduates: Alex Rogers, Austin Buono, Kate Hawkins, & Tom Hortemiller
Our approach to this research project was
highly
individual,
Mentors:
Kelly Caine
& Fuxiao Xinwith a major group component at
the end to develop our final product.
Endnote
One of the most valuable learning experiences of our research was
becoming familiar with Endnote. Endnote is a commercial reference
management software package, used to manage bibliographies and
references when writing essays and articles.
Research
Our approach to this research project was highly individual, with a
major group component at the end to develop our final product.
Web Based Health Recommender System
“Our recommendation system was designed with the goal of providing
accurate, low cost medical recommendations in countries where health
care costs are prohibitively expensive, this system can provide a free
alternative…like a second opinion.”
Extraction
while(<FL3>) {
chomp;
#this is parsing one line of the file
#escaping paring the first line, which is the field
PERSON_ID (PRIMARY KEY)
name
YEAR_OF_BIRTH (age)
if (not /^\d/) {
GENDER_COCEPT_ID (gender)
next;
SOURCE_RACE_CODE (race)
}
DRUG_CONCEPT_ID (medication taken) #split the line into each field, and put the value of
CONDITION_CONCEPT_ID (side effect) each field into an array
my @line = split(/\t/, $_);
my $Condition_Concept_ID = $line[5];
$personID = $line[2];
In the beginning, we met as a group once per week to discuss topics related to our project that
we had researched, and to determine which direction we wanted to go next. We would then
separate for the week. During this time, we would do individual research over the ideas that we
wanted to address at our next meeting. This allowed us to have multiple resources and results
when we got together to collaborate with our research. We would then discuss what we
considered
to
be
the
best
sources
of
information
to
base
our
findings
on,
and
then
repeat
this
IRB
process.
Citing sources is one of the most valuable components of a research
paper. Giving people insight to where the information originated is an
invaluable step towards further development in the future.
Endnote was an extremely useful tool in our collaboration throughout
the project. We met several times to discuss research we had found
individually, and having our sources neatly recorded made it easy to
share information with each other.
Since we knew a project like ours would likely involve sensitive data
such as patient information at some point in time, we decided to look
into the institutional review board (IRB), also known as an independent
ethics committee or ethical review board. An IRB is a committee set up
to approve, monitor, and review any research that involves humans.
They aim to maintain ethical treatment and protect the rights and
welfare of all research subjects. In the United States, the FDA along with
the Department of Health and Human Services empower IRBs to
approve, require modifications in planned research prior to approval, or
disapprove research. An IRB performs critical oversight functions for
research conducted on human subjects that are scientific and ethical.
Not all forms of research require IRB approval.
Below is a checklist that we reviewed when determining whether or not
we would need IRB approval:
In the beginning, we met as a group once per week to discuss topics
related to our project that we had researched, and to determine which
direction we wanted to go next. We would then separate for the week.
During this time, we would do individual research over the ideas that we
wanted to address at our next meeting. This allowed us to have multiple
resources and results when we got together to collaborate with our
research. We would then discuss what we considered to be the best
sources of information to base our findings on, and then repeat this
process.
1. A database of rules
2. A database of facts
3. An inference engine
We found this to be a very successful research method, and were then
able to apply what we had found to our project as a group. Reviewing
multiple resources gave us a base to build our ideas on, and allowed us
to be more knowledgeable when approaching our final task.
All of this gave us a better understanding and respect for the
importance of research.
if (not exists $line[5]) {
next;
}
$data{$personID}{'Side_Effects'}{
$Condition_Concept_ID}=1 ;
}
Recommender Systems
Provide advice to users about items they wish to purchase or examine.
The recommendations made by the system can help users search
through large amounts of information. Also, the system needs to be able
to make a connection between the user’s needs and the available items
that are in the system to be recommended. A key part to a recommender
system is the knowledge of the content that is in the system. The system
needs to be able to make a connection between the user’s needs and the
available items that are in the system to be recommended
Left Side-These rules create the fact database which supports the
Inference Engine; tasked with providing recommendations based on the
user’s input and the known facts. The fact database also contains the
information required by the explanation system, which helps users
understand their recommendation.
Right Side-A model of the entire system on a single machine
We found this to be a very successful research method, and were then able to apply what we had
found to our project as a group. Reviewing multiple resources gave us a base to build our ideas
on, and allowed us to be more knowledgeable when approaching our final task.
Is it research?
If BOTH of the following are true, your activity involves research:
-The activity is a systematic investigation, including research
development, testing and evaluation.
-The activity is designed to develop or contribute to general knowledge.
There are 4 different recommender techniques that are widely used:
1. The collaborative recommendation system has knowledge through
collaborative opinion profiles, demographic profiles and user opinions
Carbonell, J., Siekmann, J., An, A., Stefanowski, J., Ramanna, S., Butz, C., et al. (2007). Web Based Health
Recommender System Using Rough Sets, Survival Analysis and Rule-Based Expert
Systems. (p. 491). doi:10.1007/978-3-540-72530-5&#95;59.
Data Set
Orwik 2009/2010 OMOP Cup: Methods Competition
-current methods for detecting adverse drug events from observational
data are undeveloped
-helps to ensure the safety of prescription drugs
-a major global public health challenge
-there is a need for new data sources and new algorithms to respond to
the challenge
-Each challenge offers a Grand Prize
S_person: contains a unique identifier for each person along with
demographic data.
S_ObsPeriod: contains the span of time for which data is captured in the
database for a given patient.
S_DrugExp: contains periods of drug usage for each person.
S_ConditionOCC: contains records of conditions and when they were
observed.
K-nearest Neighbor
All of this gave us a better understanding and respect for the importance of research.
Does it involve human subjects?
Does my activity involve deceased individuals?
If it does, your activity requires submission to the IUPUI/Clarian IRB.
Does my research involve the FDA?
Is it a student project involving risk to human subjects?
TEMPLATE DESIGN © 2008
www.PosterPresentations.com
2. The content- based recommendation has the task of learning a
specific classification rule for each user on the basis of the user’s rating
information and the attributes of each item so that items can be
classified as likely to be interesting or not
3. The knowledge based- recommender system has an emphasis on the
user’s situation and how recommended items can meet that particular
need
4. A hybrid recommender is one that uses recommendation components
or logic of different types. For example, it may combine the methods of
collaborative and content-based recommender systems
After the extraction of the data, this information must be ran through
some sort of algorithm to distinguish what medications would be right
for a certain person. The algorithm that we chose was K Nearest
Neighbor and is the State-of- the- art algorithm used in recommender
systems. With this algorithm we are capable of finding similarities
between patients and predicting the best medication for them. Using
this algorithm also allowed use to create this type of vision with the data
we extracted:
Health
Recommender
Recommender
SystemsSystem
Undergraduates: Alex Rogers, Austin Buono, Kate Hawkins, & Tom Hortemiller
Provide advice to users about items they wish to purchase
examine.
Mentors: Kelly or
Caine
& Fuxiao Xin The recommendations made by the
system can help users search through large amounts of information. Also, the system needs to be able to
Endnote
Research
Web Based Health Recommender System
Extraction
make a connection between the user’s needs and the available items that are in the system to be
recommended. A key part to a recommender system is the knowledge of the content that is in the system.
The system needs to be able to make a connection between the user’s needs and the available items that are
in the system to be recommended
One of the most valuable learning experiences of our research was
becoming familiar with Endnote. Endnote is a commercial reference
management software package, used to manage bibliographies and
references when writing essays and articles.
Citing sources is one of the most valuable components of a research
paper. Giving people insight to where the information originated is an
invaluable step towards further development in the future.
Our approach to this research project was highly individual, with a
major group component at the end to develop our final product.
In the beginning, we met as a group once per week to discuss topics
related to our project that we had researched, and to determine which
direction we wanted to go next. We would then separate for the week.
During this time, we would do individual research over the ideas that we
wanted to address at our next meeting. This allowed us to have multiple
resources and results when we got together to collaborate with our
research. We would then discuss what we considered to be the best
sources of information to base our findings on, and then repeat this
process.
“Our recommendation system was designed with the goal of providing
accurate, low cost medical recommendations in countries where health
care costs are prohibitively expensive, this system can provide a free
alternative…like a second opinion.”
1. A database of rules
2. A database of facts
3. An inference engine
while(<FL3>) {
chomp;
#this is parsing one line of the file
#escaping paring the first line, which is the field
PERSON_ID (PRIMARY KEY)
name
YEAR_OF_BIRTH (age)
if (not /^\d/) {
GENDER_COCEPT_ID (gender)
next;
SOURCE_RACE_CODE (race)
}
DRUG_CONCEPT_ID (medication taken) #split the line into each field, and put the value of
CONDITION_CONCEPT_ID (side effect) each field into an array
my @line = split(/\t/, $_);
my $Condition_Concept_ID = $line[5];
$personID = $line[2];
There are 4 different recommender techniques that are widely used:
1. The collaborative
recommendation
system
has
knowledge
through
collaborative
opinion
profiles,
IRB
demographic profiles and user opinions
Endnote was an extremely useful tool in our collaboration throughout
the project. We met several times to discuss research we had found
individually, and having our sources neatly recorded made it easy to
share information with each other.
Since we knew a project like ours would likely involve sensitive data
such as patient information at some point in time, we decided to look
into the institutional review board (IRB), also known as an independent
ethics committee or ethical review board. An IRB is a committee set up
to approve, monitor, and review any research that involves humans.
They aim to maintain ethical treatment and protect the rights and
welfare of all research subjects. In the United States, the FDA along with
the Department of Health and Human Services empower IRBs to
approve, require modifications in planned research prior to approval, or
disapprove research. An IRB performs critical oversight functions for
research conducted on human subjects that are scientific and ethical.
Not all forms of research require IRB approval.
Below is a checklist that we reviewed when determining whether or not
we would need IRB approval:
We found this to be a very successful research method, and were then
able to apply what we had found to our project as a group. Reviewing
multiple resources gave us a base to build our ideas on, and allowed us
to be more knowledgeable when approaching our final task.
if (not exists $line[5]) {
next;
}
$data{$personID}{'Side_Effects'}{
$Condition_Concept_ID}=1 ;
All of this gave us a better understanding and respect for the
importance of research.
}
Recommender Systems
Provide advice to users about items they wish to purchase or examine.
The recommendations made by the system can help users search
through large amounts of information. Also, the system needs to be able
to make a connection between the user’s needs and the available items
that are in the system to be recommended. A key part to a recommender
system is the knowledge of the content that is in the system. The system
needs to be able to make a connection between the user’s needs and the
available items that are in the system to be recommended
Left Side-These rules create the fact database which supports the
Inference Engine; tasked with providing recommendations based on the
user’s input and the known facts. The fact database also contains the
information required by the explanation system, which helps users
understand their recommendation.
Right Side-A model of the entire system on a single machine
2. The content- based recommendation has the task of learning a specific classification rule for each user
on the basis of the user’s rating information and the attributes of each item so that items can be classified as
likely to be interesting or not
Is it research?
If BOTH of the following are true, your activity involves research:
-The activity is a systematic investigation, including research
development, testing and evaluation.
-The activity is designed to develop or contribute to general knowledge.
There are 4 different recommender techniques that are widely used:
1. The collaborative recommendation system has knowledge through
collaborative opinion profiles, demographic profiles and user opinions
Carbonell, J., Siekmann, J., An, A., Stefanowski, J., Ramanna, S., Butz, C., et al. (2007). Web Based Health
Recommender System Using Rough Sets, Survival Analysis and Rule-Based Expert
Systems. (p. 491). doi:10.1007/978-3-540-72530-5&#95;59.
Data Set
Orwik 2009/2010 OMOP Cup: Methods Competition
-current methods for detecting adverse drug events from observational
data are undeveloped
-helps to ensure the safety of prescription drugs
-a major global public health challenge
-there is a need for new data sources and new algorithms to respond to
the challenge
-Each challenge offers a Grand Prize
S_person: contains a unique identifier for each person along with
demographic data.
S_ObsPeriod: contains the span of time for which data is captured in the
database for a given patient.
S_DrugExp: contains periods of drug usage for each person.
S_ConditionOCC: contains records of conditions and when they were
observed.
K-nearest Neighbor
3. The knowledge based- recommender system has an emphasis on the user’s situation and how
recommended items can meet that particular need
Does it involve human subjects?
Does my activity involve deceased individuals?
If it does, your activity requires submission to the IUPUI/Clarian IRB.
2. The content- based recommendation has the task of learning a
specific classification rule for each user on the basis of the user’s rating
information and the attributes of each item so that items can be
classified as likely to be interesting or not
3. The knowledge based- recommender system has an emphasis on the
user’s situation and how recommended items can meet that particular
need
After the extraction of the data, this information must be ran through
some sort of algorithm to distinguish what medications would be right
for a certain person. The algorithm that we chose was K Nearest
Neighbor and is the State-of- the- art algorithm used in recommender
systems. With this algorithm we are capable of finding similarities
between patients and predicting the best medication for them. Using
this algorithm also allowed use to create this type of vision with the data
we extracted:
4. A hybrid recommender is one that uses recommendation components or logic of different types. For
example, it may combine the methods of collaborative and content-based recommender systems
Does my research involve the FDA?
Is it a student project involving risk to human subjects?
TEMPLATE DESIGN © 2008
www.PosterPresentations.com
4. A hybrid recommender is one that uses recommendation components
or logic of different types. For example, it may combine the methods of
collaborative and content-based recommender systems
Recommender
System
WebHealth
Based Health
Recommender
System
Undergraduates: Alex Rogers, Austin Buono, Kate Hawkins, & Tom Hortemiller
Mentors: Kelly Caine & Fuxiao Xin
“Our recommendation system was designed with the
goal of providing
accurate,
low
cost
medical
Endnote
Research
recommendations in countries where health care costs
are prohibitively expensive, this system can provide a
free alternative…like a second opinion.”
One of the most valuable learning experiences of our research was
becoming familiar with Endnote. Endnote is a commercial reference
management software package, used to manage bibliographies and
references when writing essays and articles.
Citing sources is one of the most valuable components of a research
paper. Giving people insight to where the information originated is an
invaluable step towards further development in the future.
1.A database of rules
2.A database of facts
IRB engine
3.An inference
Endnote was an extremely useful tool in our collaboration throughout
the project. We met several times to discuss research we had found
individually, and having our sources neatly recorded made it easy to
share information with each other.
Since we knew a project like ours would likely involve sensitive data
such as patient information at some point in time, we decided to look
into the institutional review board (IRB), also known as an independent
ethics committee or ethical review board. An IRB is a committee set up
to approve, monitor, and review any research that involves humans.
They aim to maintain ethical treatment and protect the rights and
welfare of all research subjects. In the United States, the FDA along with
the Department of Health and Human Services empower IRBs to
approve, require modifications in planned research prior to approval, or
disapprove research. An IRB performs critical oversight functions for
research conducted on human subjects that are scientific and ethical.
Not all forms of research require IRB approval.
Below is a checklist that we reviewed when determining whether or not
we would need IRB approval:
Is it research?
If BOTH of the following are true, your activity involves research:
-The activity is a systematic investigation, including research
development, testing and evaluation.
-The activity is designed to develop or contribute to general knowledge.
Does it involve human subjects?
Does my activity involve deceased individuals?
If it does, your activity requires submission to the IUPUI/Clarian IRB.
Does my research involve the FDA?
Is it a student project involving risk to human subjects?
TEMPLATE DESIGN © 2008
www.PosterPresentations.com
Our approach to this research project was highly individual, with a
major group component at the end to develop our final product.
In the beginning, we met as a group once per week to discuss topics
related to our project that we had researched, and to determine which
direction we wanted to go next. We would then separate for the week.
During this time, we would do individual research over the ideas that we
wanted to address at our next meeting. This allowed us to have multiple
resources and results when we got together to collaborate with our
research. We would then discuss what we considered to be the best
sources of information to base our findings on, and then repeat this
process.
Web Based Health Recommender System
“Our recommendation system was designed with the goal of providing
accurate, low cost medical recommendations in countries where health
care costs are prohibitively expensive, this system can provide a free
alternative…like a second opinion.”
1. A database of rules
2. A database of facts
3. An inference engine
We found this to be a very successful research method, and were then
able to apply what we had found to our project as a group. Reviewing
multiple resources gave us a base to build our ideas on, and allowed us
to be more knowledgeable when approaching our final task.
2. The content- based recommendation has the task of learning a
specific classification rule for each user on the basis of the user’s rating
information and the attributes of each item so that items can be
classified as likely to be interesting or not
3. The knowledge based- recommender system has an emphasis on the
user’s situation and how recommended items can meet that particular
need
4. A hybrid recommender is one that uses recommendation components
or logic of different types. For example, it may combine the methods of
collaborative and content-based recommender systems
Left Side-These rules create the
fact database which supports the
Inference Engine; tasked with
providing recommendations based
on the user’s input and the known
facts. The fact database also
contains the information required
by the explanation system, which
helps users understand their
recommendation.
Right Side-A model
of the entire
K-nearest Neighbor
system on a single machine
}
Recommender Systems
There are 4 different recommender techniques that are widely used:
1. The collaborative recommendation system has knowledge through
collaborative opinion profiles, demographic profiles and user opinions
while(<FL3>) {
chomp;
#this is parsing one line of the file
#escaping paring the first line, which is the field
PERSON_ID (PRIMARY KEY)
name
YEAR_OF_BIRTH (age)
if (not /^\d/) {
GENDER_COCEPT_ID (gender)
next;
SOURCE_RACE_CODE (race)
}
DRUG_CONCEPT_ID (medication taken) #split the line into each field, and put the value of
CONDITION_CONCEPT_ID (side effect) each field into an array
my @line = split(/\t/, $_);
my $Condition_Concept_ID = $line[5];
$personID = $line[2];
if (not exists $line[5]) {
next;
}
$data{$personID}{'Side_Effects'}{
$Condition_Concept_ID}=1 ;
All of this gave us a better understanding and respect for the
importance of research.
Provide advice to users about items they wish to purchase or examine.
The recommendations made by the system can help users search
through large amounts of information. Also, the system needs to be able
to make a connection between the user’s needs and the available items
that are in the system to be recommended. A key part to a recommender
system is the knowledge of the content that is in the system. The system
needs to be able to make a connection between the user’s needs and the
available items that are in the system to be recommended
Extraction
Left Side-These rules create the fact database which supports the
Inference Engine; tasked with providing recommendations based on the
user’s input and the known facts. The fact database also contains the
information required by the explanation system, which helps users
understand their recommendation.
Right Side-A model of the entire system on a single machine
Carbonell, J., Siekmann, J., An, A., Stefanowski, J., Ramanna, S., Butz, C., et al. (2007). Web Based Health
Recommender System Using Rough Sets, Survival Analysis and Rule-Based Expert
Systems. (p. 491). doi:10.1007/978-3-540-72530-5&#95;59.
Data Set
Orwik 2009/2010 OMOP Cup: Methods Competition
-current methods for detecting adverse drug events from observational
data are undeveloped
-helps to ensure the safety of prescription drugs
-a major global public health challenge
-there is a need for new data sources and new algorithms to respond to
the challenge
-Each challenge offers a Grand Prize
S_person: contains a unique identifier for each person along with
demographic data.
S_ObsPeriod: contains the span of time for which data is captured in the
database for a given patient.
S_DrugExp: contains periods of drug usage for each person.
S_ConditionOCC: contains records of conditions and when they were
observed.
After the extraction of the data, this information must be ran through
some sort of algorithm to distinguish what medications would be right
for a certain person. The algorithm that we chose was K Nearest
Neighbor and is the State-of- the- art algorithm used in recommender
systems. With this algorithm we are capable of finding similarities
between patients and predicting the best medication for them. Using
this algorithm also allowed use to create this type of vision with the data
we extracted:
Carbonell, J., Siekmann, J., An, A., Stefanowski, J., Ramanna, S., Butz, C., et al. (2007).
Web Based Health Recommender System Using Rough Sets, Survival Analysis
and Rule-Based Expert Systems. (p. 491). doi:10.1007/978-3-540-72530-5&#95;59.
Research Conclusions
-Recommender systems have been successful in consumer platforms such as
Amazon, Netflix, and Facebook
-Recommender systems have been used in the Health field
-Use has been limited by confidentiality and reliance on a single system over
human medicine
-State of the art involves the “tagging” technique and the use of an algorithm such
as “k-nearest neighbor”
POSTER TEMPLATE BY:
www.PosterPresentations.com
Health Recommender System
Data Set
Undergraduates: Alex Rogers, Austin Buono, Kate Hawkins, & Tom Hortemiller
Mentors: Kelly Caine & Fuxiao Xin
Endnote
Orwik 2009/2010
OMOP Cup: Methods
Research
Web Based Health Competition
Recommender System
-current methods for detecting adverse drug events from observational
data are undeveloped
-helps to ensure the safety of prescription drugs
-a major global public health challenge
-there is a need for new data sources and new algorithms to respond to
the challenge
-Each challenge offers a Grand Prize
S_person: contains
a unique
identifier for each person along with
Recommender
Systems
demographic data.
S_ObsPeriod: contains the span of time for which data is captured in the
database for a given patient.
S_DrugExp: contains periods of drug usage for each
Data Set person.
S_ConditionOCC: contains records of conditions and when they were
observed.
One of the most valuable learning experiences of our research was
becoming familiar with Endnote. Endnote is a commercial reference
management software package, used to manage bibliographies and
references when writing essays and articles.
Citing sources is one of the most valuable components of a research
paper. Giving people insight to where the information originated is an
invaluable step towards further development in the future.
Endnote was an extremely useful tool in our collaboration throughout
the project. We met several times to discuss research we had found
individually, and having our sources neatly recorded made it easy to
share information with each other.
IRB
Since we knew a project like ours would likely involve sensitive data
such as patient information at some point in time, we decided to look
into the institutional review board (IRB), also known as an independent
ethics committee or ethical review board. An IRB is a committee set up
to approve, monitor, and review any research that involves humans.
They aim to maintain ethical treatment and protect the rights and
welfare of all research subjects. In the United States, the FDA along with
the Department of Health and Human Services empower IRBs to
approve, require modifications in planned research prior to approval, or
disapprove research. An IRB performs critical oversight functions for
research conducted on human subjects that are scientific and ethical.
Not all forms of research require IRB approval.
Below is a checklist that we reviewed when determining whether or not
we would need IRB approval:
Is it research?
If BOTH of the following are true, your activity involves research:
-The activity is a systematic investigation, including research
development, testing and evaluation.
-The activity is designed to develop or contribute to general knowledge.
Does it involve human subjects?
Does my activity involve deceased individuals?
If it does, your activity requires submission to the IUPUI/Clarian IRB.
Does my research involve the FDA?
Is it a student project involving risk to human subjects?
TEMPLATE DESIGN © 2008
www.PosterPresentations.com
Our approach to this research project was highly individual, with a
major group component at the end to develop our final product.
In the beginning, we met as a group once per week to discuss topics
related to our project that we had researched, and to determine which
direction we wanted to go next. We would then separate for the week.
During this time, we would do individual research over the ideas that we
wanted to address at our next meeting. This allowed us to have multiple
resources and results when we got together to collaborate with our
research. We would then discuss what we considered to be the best
sources of information to base our findings on, and then repeat this
process.
“Our recommendation system was designed with the goal of providing
accurate, low cost medical recommendations in countries where health
care costs are prohibitively expensive, this system can provide a free
alternative…like a second opinion.”
1. A database of rules
2. A database of facts
3. An inference engine
Extraction
while(<FL3>) {
chomp;
#this is parsing one line of the file
#escaping paring the first line, which is the field
PERSON_ID (PRIMARY KEY)
name
YEAR_OF_BIRTH (age)
if (not /^\d/) {
GENDER_COCEPT_ID (gender)
next;
SOURCE_RACE_CODE (race)
}
DRUG_CONCEPT_ID (medication taken) #split the line into each field, and put the value of
CONDITION_CONCEPT_ID (side effect) each field into an array
my @line = split(/\t/, $_);
my $Condition_Concept_ID = $line[5];
$personID = $line[2];
We found this to be a very successful research method, and were then
able to apply what we had found to our project as a group. Reviewing
multiple resources gave us a base to build our ideas on, and allowed us
to be more knowledgeable when approaching our final task.
All of this gave us a better understanding and respect for the
importance of research.
Provide advice to users about items they wish to purchase or examine.
The recommendations made by the system can help users search
through large amounts of information. Also, the system needs to be able
to make a connection between the user’s needs and the available items
that are in the system to be recommended. A key part to a recommender
system is the knowledge of the content that is in the system. The system
needs to be able to make a connection between the user’s needs and the
available items that are in the system to be recommended
There are 4 different recommender techniques that are widely used:
1. The collaborative recommendation system has knowledge through
collaborative opinion profiles, demographic profiles and user opinions
2. The content- based recommendation has the task of learning a
specific classification rule for each user on the basis of the user’s rating
information and the attributes of each item so that items can be
classified as likely to be interesting or not
3. The knowledge based- recommender system has an emphasis on the
user’s situation and how recommended items can meet that particular
need
4. A hybrid recommender is one that uses recommendation components
or logic of different types. For example, it may combine the methods of
collaborative and content-based recommender systems
if (not exists $line[5]) {
next;
}
$data{$personID}{'Side_Effects'}{
$Condition_Concept_ID}=1 ;
}
Left Side-These rules create the fact database which supports the
Inference Engine; tasked with providing recommendations based on the
user’s input and the known facts. The fact database also contains the
information required by the explanation system, which helps users
understand their recommendation.
Right Side-A model of the entire system on a single machine
Carbonell, J., Siekmann, J., An, A., Stefanowski, J., Ramanna, S., Butz, C., et al. (2007). Web Based Health
Recommender System Using Rough Sets, Survival Analysis and Rule-Based Expert
Systems. (p. 491). doi:10.1007/978-3-540-72530-5&#95;59.
Orwik 2009/2010 OMOP Cup: Methods Competition
-current methods for detecting adverse drug events from observational
data are undeveloped
-helps to ensure the safety of prescription drugs
-a major global public health challenge
-there is a need for new data sources and new algorithms to respond to
the challenge
-Each challenge offers a Grand Prize
S_person: contains a unique identifier for each person along with
demographic data.
S_ObsPeriod: contains the span of time for which data is captured in the
database for a given patient.
S_DrugExp: contains periods of drug usage for each person.
S_ConditionOCC: contains records of conditions and when they were
observed.
K-nearest Neighbor
After the extraction of the data, this information must be ran through
some sort of algorithm to distinguish what medications would be right
for a certain person. The algorithm that we chose was K Nearest
Neighbor and is the State-of- the- art algorithm used in recommender
systems. With this algorithm we are capable of finding similarities
between patients and predicting the best medication for them. Using
this algorithm also allowed use to create this type of vision with the data
we extracted:
my %data = {};
while(<FL3>) {
Health
Recommender
System
Extraction
chomp;
#specify the file name
Undergraduates: Alex Rogers, Austin Buono, Kate Hawkins, & Tom Hortemiller
my $file = 'S_person.tab' ;
Mentors: Kelly Caine & Fuxiao Xin
#this is parsing one line of the file
my $file2 = 'S_DrugExp.tab';
#escaping paring the first line, which is the field name
my $file3 = 'S_ConditionOCC.tab';
Endnote
Research
Web Based Health Recommender System
Extraction
ifOne(not
/^\d/)
{
of the most valuable learning experiences of our research was
Our approach to this research project was highly individual, with a
“Our recommendation system was designed with the goal of providing
#open file
becoming familiar with Endnote. Endnote is a commercial reference
major group component at the end to develop our final product.
accurate, low cost medical recommendations in countries where health
next;software package, used to manage bibliographies and
management
care costs are prohibitively expensive, this system
can provide a free
open(FL,
$file) || die "Can not open the $file\n";
references when writing essays and articles.
In the beginning, we met as a group once per week to discuss topics
alternative…like a second opinion.”
related to our project that we had researched, and to determine which
open(FL2, $file2) || die "Can not open the $file2\n";
}Citing sources is one of the most valuable components of a research
direction we wanted to go next. We would then separate for the week.
1. A database of rules
open(FL3, $file3) || die "Can not open the $file3\n";
paper. Giving people insight to where the information originated is an
During this time, we would do individual research over the ideas that we
2. A database of facts
#split
thetowards
linefurther
into
eachin field,
value
of ateach
field This
into
invaluable step
development
the future. and put thewanted
to address
our next meeting.
allowed us to have multiple
3. An inference engine
resources and results when we got together to collaborate with our
research. We would then discuss what we considered to be the best
an array
sources of information to base our findings on, and then repeat this
process.
my
= split(/\t/,
$_);
Endnote@line
was an extremely
useful tool in our
collaboration throughout
the project. We met several times to discuss research we had found
We found this to be a very successful research method, and were then
individually,
and having our sources neatly recorded made
it $line[5];
easy to
able to apply what we had found to our project as a group. Reviewing
my
$Condition_Concept_ID
=
share information with each other.
multiple resources gave us a base to build our ideas on, and allowed us
to be more knowledgeable when approaching our final task.
IRB
$personID = $line[2];
while(<FL3>) {
chomp;
#this is parsing one line of the file
#escaping paring the first line, which is the field
PERSON_ID (PRIMARY KEY)
name
YEAR_OF_BIRTH (age)
if (not /^\d/) {
GENDER_COCEPT_ID (gender)
next;
SOURCE_RACE_CODE (race)
}
DRUG_CONCEPT_ID (medication taken) #split the line into each field, and put the value of
CONDITION_CONCEPT_ID (side effect) each field into an array
my @line = split(/\t/, $_);
my $Condition_Concept_ID = $line[5];
$personID = $line[2];
Since we knew a project like ours would likely involve sensitive data
such as patient information at some point in time, we decided to look
into the institutional review board (IRB), also known as an independent
ethics committee or ethical review board. An IRB is a committee set up
to approve, monitor, and review any research that involves humans.
They aim to maintain ethical treatment and protect the rights and
welfare of all research subjects. In the United States, the FDA along with
the Department of Health and Human Services empower IRBs to
approve, require modifications in planned research prior to approval, or
disapprove research. An IRB performs critical oversight functions for
research conducted on human subjects that are scientific and ethical.
Not all forms of research require IRB approval.
Below is a checklist that we reviewed when determining whether or not
we would need IRB approval:
if (not exists $line[5]) {
next;
}
$data{$personID}{'Side_Effects'}{
$Condition_Concept_ID}=1 ;
} PERSON_ID (PRIMARY KEY)
Is it research?
If BOTH of the following are true, your activity involves research:
-The activity is a systematic investigation, including research
development, testing and evaluation.
-The activity is designed to develop or contribute to general knowledge.
All of this gave us a better understanding and respect for the
importance of research.
www.PosterPresentations.com
}
Recommender Systems
Provide advice to users about items they wish to purchase or examine.
The recommendations made by the system can help users search
through large amounts of information. Also, the system needs to be able
to make a connection between the user’s needs and the available items
that are in the system to be recommended. A key part to a recommender
system is the knowledge of the content that is in the system. The system
needs to be able to make a connection between the user’s needs and the
available items that are in the system to be recommended
There are 4 different recommender techniques that are widely used:
1. The collaborative recommendation system has knowledge through
collaborative opinion profiles, demographic profiles and user opinions
2. The content- based recommendation has the task of learning a
specific classification rule for each user on the basis of the user’s rating
information and the attributes of each item so that items can be
classified as likely to be interesting or not
YEAR_OF_BIRTH (age)
Does
it involve human subjects?
GENDER_COCEPT_ID
(gender) 3. The knowledge based- recommender system has an emphasis on the
user’s situation and how recommended items can meet that particular
Does
my
activity
involve
deceased
individuals?
SOURCE_RACE_CODE
(race) need
If it does, your activity requires submission to the IUPUI/Clarian IRB.
4. A hybrid
recommender is one that uses recommendation components
DRUG_CONCEPT_ID
(medication
taken)
Does my research involve the FDA?
or logic of different types. For example, it may combine the methods of
and content-based recommender systems
CONDITION_CONCEPT_ID
(sidecollaborative
effect)
Is it a student project involving risk to human subjects?
TEMPLATE DESIGN © 2008
if (not exists $line[5]) {
next;
}
$data{$personID}{'Side_Effects'}{
$Condition_Concept_ID}=1 ;
Left Side-These rules create the fact database which supports the
Inference Engine; tasked with providing recommendations based on the
user’s input and the known facts. The fact database also contains the
information required by the explanation system, which helps users
understand their recommendation.
Right Side-A model of the entire system on a single machine
Carbonell, J., Siekmann, J., An, A., Stefanowski, J., Ramanna, S., Butz, C., et al. (2007). Web Based Health
Recommender System Using Rough Sets, Survival Analysis and Rule-Based Expert
Systems. (p. 491). doi:10.1007/978-3-540-72530-5&#95;59.
Data Set
Orwik 2009/2010 OMOP Cup: Methods Competition
-current methods for detecting adverse drug events from observational
data are undeveloped
-helps to ensure the safety of prescription drugs
-a major global public health challenge
-there is a need for new data sources and new algorithms to respond to
the challenge
-Each challenge offers a Grand Prize
S_person: contains a unique identifier for each person along with
demographic data.
S_ObsPeriod: contains the span of time for which data is captured in the
database for a given patient.
S_DrugExp: contains periods of drug usage for each person.
S_ConditionOCC: contains records of conditions and when they were
observed.
K-nearest Neighbor
After the extraction of the data, this information must be ran through
some sort of algorithm to distinguish what medications would be right
for a certain person. The algorithm that we chose was K Nearest
Neighbor and is the State-of- the- art algorithm used in recommender
systems. With this algorithm we are capable of finding similarities
between patients and predicting the best medication for them. Using
this algorithm also allowed use to create this type of vision with the data
we extracted:
Health Recommender System
K-nearest Neighbor
Undergraduates: Alex Rogers, Austin Buono, Kate Hawkins, & Tom Hortemiller
Mentors: Kelly Caine & Fuxiao Xin
Endnote
After the extraction of
the
data,
this
information
must
be
ran
through
Research
Web Based Health Recommender System
some sort of algorithm to distinguish what medications would be right
for a certain person. The algorithm that we chose was K Nearest
Neighbor and is the State-of- the- art algorithm used in recommender
systems. With this algorithm we are capable of finding similarities
between patients and predicting the best medication for them. Using this
algorithm also allowed use to create this type of vision with the data we
extracted:
One of the most valuable learning experiences of our research was
becoming familiar with Endnote. Endnote is a commercial reference
management software package, used to manage bibliographies and
references when writing essays and articles.
Citing sources is one of the most valuable components of a research
paper. Giving people insight to where the information originated is an
invaluable step towards further development in the future.
Endnote was an extremely useful tool in our collaboration throughout
the project. We met several times to discuss research we had found
individually, and having our sources neatly recorded made it easy to
share information with each other.
IRB
Since we knew a project like ours would likely involve sensitive data
such as patient information at some point in time, we decided to look
into the institutional review board (IRB), also known as an independent
ethics committee or ethical review board. An IRB is a committee set up
to approve, monitor, and review any research that involves humans.
They aim to maintain ethical treatment and protect the rights and
welfare of all research subjects. In the United States, the FDA along with
the Department of Health and Human Services empower IRBs to
approve, require modifications in planned research prior to approval, or
disapprove research. An IRB performs critical oversight functions for
research conducted on human subjects that are scientific and ethical.
Not all forms of research require IRB approval.
Below is a checklist that we reviewed when determining whether or not
we would need IRB approval:
Is it research?
If BOTH of the following are true, your activity involves research:
-The activity is a systematic investigation, including research
development, testing and evaluation.
-The activity is designed to develop or contribute to general knowledge.
Does it involve human subjects?
Does my activity involve deceased individuals?
If it does, your activity requires submission to the IUPUI/Clarian IRB.
Does my research involve the FDA?
Is it a student project involving risk to human subjects?
TEMPLATE DESIGN © 2008
www.PosterPresentations.com
Our approach to this research project was highly individual, with a
major group component at the end to develop our final product.
In the beginning, we met as a group once per week to discuss topics
related to our project that we had researched, and to determine which
direction we wanted to go next. We would then separate for the week.
During this time, we would do individual research over the ideas that we
wanted to address at our next meeting. This allowed us to have multiple
resources and results when we got together to collaborate with our
research. We would then discuss what we considered to be the best
sources of information to base our findings on, and then repeat this
process.
“Our recommendation system was designed with the goal of providing
accurate, low cost medical recommendations in countries where health
care costs are prohibitively expensive, this system can provide a free
alternative…like a second opinion.”
1. A database of rules
2. A database of facts
3. An inference engine
Extraction
while(<FL3>) {
chomp;
#this is parsing one line of the file
#escaping paring the first line, which is the field
PERSON_ID (PRIMARY KEY)
name
YEAR_OF_BIRTH (age)
if (not /^\d/) {
GENDER_COCEPT_ID (gender)
next;
SOURCE_RACE_CODE (race)
}
DRUG_CONCEPT_ID (medication taken) #split the line into each field, and put the value of
CONDITION_CONCEPT_ID (side effect) each field into an array
my @line = split(/\t/, $_);
my $Condition_Concept_ID = $line[5];
$personID = $line[2];
We found this to be a very successful research method, and were then
able to apply what we had found to our project as a group. Reviewing
multiple resources gave us a base to build our ideas on, and allowed us
to be more knowledgeable when approaching our final task.
All of this gave us a better understanding and respect for the
importance of research.
if (not exists $line[5]) {
next;
}
$data{$personID}{'Side_Effects'}{
$Condition_Concept_ID}=1 ;
}
Recommender Systems
Provide advice to users about items they wish to purchase or examine.
The recommendations made by the system can help users search
through large amounts of information. Also, the system needs to be able
to make a connection between the user’s needs and the available items
that are in the system to be recommended. A key part to a recommender
system is the knowledge of the content that is in the system. The system
needs to be able to make a connection between the user’s needs and the
available items that are in the system to be recommended
There are 4 different recommender techniques that are widely used:
1. The collaborative recommendation system has knowledge through
collaborative opinion profiles, demographic profiles and user opinions
2. The content- based recommendation has the task of learning a
specific classification rule for each user on the basis of the user’s rating
information and the attributes of each item so that items can be
classified as likely to be interesting or not
3. The knowledge based- recommender system has an emphasis on the
user’s situation and how recommended items can meet that particular
need
4. A hybrid recommender is one that uses recommendation components
or logic of different types. For example, it may combine the methods of
collaborative and content-based recommender systems
Left Side-These rules create the fact database which supports the
Inference Engine; tasked with providing recommendations based on the
user’s input and the known facts. The fact database also contains the
information required by the explanation system, which helps users
understand their recommendation.
Right Side-A model of the entire system on a single machine
Carbonell, J., Siekmann, J., An, A., Stefanowski, J., Ramanna, S., Butz, C., et al. (2007). Web Based Health
Recommender System Using Rough Sets, Survival Analysis and Rule-Based Expert
Systems. (p. 491). doi:10.1007/978-3-540-72530-5&#95;59.
Data Set
Orwik 2009/2010 OMOP Cup: Methods Competition
-current methods for detecting adverse drug events from observational
data are undeveloped
-helps to ensure the safety of prescription drugs
-a major global public health challenge
-there is a need for new data sources and new algorithms to respond to
the challenge
-Each challenge offers a Grand Prize
S_person: contains a unique identifier for each person along with
demographic data.
S_ObsPeriod: contains the span of time for which data is captured in the
database for a given patient.
S_DrugExp: contains periods of drug usage for each person.
S_ConditionOCC: contains records of conditions and when they were
observed.
K-nearest Neighbor
After the extraction of the data, this information must be ran through
some sort of algorithm to distinguish what medications would be right
for a certain person. The algorithm that we chose was K Nearest
Neighbor and is the State-of- the- art algorithm used in recommender
systems. With this algorithm we are capable of finding similarities
between patients and predicting the best medication for them. Using
this algorithm also allowed use to create this type of vision with the data
we extracted:
Possibilities for the Future
•Access to a real data set could allow for further data extraction leading to
more accurate predictions
•More time would allow for more detailed programming to achieve a higher
functioning recommender system
•A higher functioning health recommender system could be helpful in the
future of healthcare and the quality of care available to everyone
POSTER TEMPLATE BY:
www.PosterPresentations.com
Endnote
One of the most valuable learning experiences of our research was
becoming familiar with Endnote. Endnote is a commercial reference
management software package, used to manage bibliographies and
references when writing essays and articles.
Citing sources is one of the most valuable components of a research
paper. Giving people insight to where the information originated is an
invaluable step towards further development in the future.
Endnote was an extremely useful tool in our collaboration throughout
the project. We met several times to discuss research we had found
individually, and having our sources neatly recorded made it easy to
share information with each other.
IRB
Since we knew a project like ours would likely involve sensitive data
such as patient information at some point in time, we decided to look
into the institutional review board (IRB), also known as an independent
ethics committee or ethical review board. An IRB is a committee set up
to approve, monitor, and review any research that involves humans.
They aim to maintain ethical treatment and protect the rights and
welfare of all research subjects. In the United States, the FDA along with
the Department of Health and Human Services empower IRBs to
approve, require modifications in planned research prior to approval, or
disapprove research. An IRB performs critical oversight functions for
research conducted on human subjects that are scientific and ethical.
Not all forms of research require IRB approval.
Below is a checklist that we reviewed when determining whether or not
we would need IRB approval:
Is it research?
If BOTH of the following are true, your activity involves research:
-The activity is a systematic investigation, including research
development, testing and evaluation.
-The activity is designed to develop or contribute to general knowledge.
Does it involve human subjects?
Does my activity involve deceased individuals?
If it does, your activity requires submission to the IUPUI/Clarian IRB.
Does my research involve the FDA?
Is it a student project involving risk to human subjects?
TEMPLATE DESIGN © 2008
www.PosterPresentations.com
Research
Our approach to this research project was highly individual, with a
major group component at the end to develop our final product.
In the beginning, we met as a group once per week to discuss topics
related to our project that we had researched, and to determine which
direction we wanted to go next. We would then separate for the week.
During this time, we would do individual research over the ideas that we
wanted to address at our next meeting. This allowed us to have multiple
resources and results when we got together to collaborate with our
research. We would then discuss what we considered to be the best
sources of information to base our findings on, and then repeat this
process.
Web Based Health Recommender System
“Our recommendation system was designed with the goal of providing
accurate, low cost medical recommendations in countries where health
care costs are prohibitively expensive, this system can provide a free
alternative…like a second opinion.”
1. A database of rules
2. A database of facts
3. An inference engine
Extraction
while(<FL3>) {
chomp;
#this is parsing one line of the file
#escaping paring the first line, which is the field
name
PERSON_ID (PRIMARY KEY)
if (not /^\d/) {
YEAR_OF_BIRTH (age)
next;
GENDER_COCEPT_ID (gender)
}
SOURCE_RACE_CODE (race)
#split the line into each field, and put the value of
DRUG_CONCEPT_ID (medication taken) each field into an array
CONDITION_CONCEPT_ID (side effect) my @line = split(/\t/, $_);
my $Condition_Concept_ID = $line[5];
$personID = $line[2];
We found this to be a very successful research method, and were then
able to apply what we had found to our project as a group. Reviewing
multiple resources gave us a base to build our ideas on, and allowed us
to be more knowledgeable when approaching our final task.
if (not exists $line[5]) {
next;
}
$data{$personID}{'Side_Effects'}{
$Condition_Concept_ID}=1 ;
All of this gave us a better understanding and respect for the
importance of research.
}
Recommender Systems
Provide advice to users about items they wish to purchase or examine.
The recommendations made by the system can help users search
through large amounts of information. Also, the system needs to be able
to make a connection between the user’s needs and the available items
that are in the system to be recommended. A key part to a recommender
system is the knowledge of the content that is in the system. The system
needs to be able to make a connection between the user’s needs and the
available items that are in the system to be recommended
There are 4 different recommender techniques that are widely used:
1. The collaborative recommendation system has knowledge through
collaborative opinion profiles, demographic profiles and user opinions
2. The content- based recommendation has the task of learning a
specific classification rule for each user on the basis of the user’s rating
information and the attributes of each item so that items can be
classified as likely to be interesting or not
3. The knowledge based- recommender system has an emphasis on the
user’s situation and how recommended items can meet that particular
need
4. A hybrid recommender is one that uses recommendation components
or logic of different types. For example, it may combine the methods of
collaborative and content-based recommender systems
Left Side-These rules create the fact database which supports the
Inference Engine; tasked with providing recommendations based on the
user’s input and the known facts. The fact database also contains the
information required by the explanation system, which helps users
understand their recommendation.
Right Side-A model of the entire system on a single machine
Carbonell, J., Siekmann, J., An, A., Stefanowski, J., Ramanna, S., Butz, C., et al. (2007). Web Based Health
Recommender System Using Rough Sets, Survival Analysis and Rule-Based Expert
Systems. (p. 491). doi:10.1007/978-3-540-72530-5&#95;59.
Data Set
Orwik 2009/2010 OMOP Cup: Methods Competition
-current methods for detecting adverse drug events from observational
data are undeveloped
-helps to ensure the safety of prescription drugs
-a major global public health challenge
-there is a need for new data sources and new algorithms to respond to
the challenge
-Each challenge offers a Grand Prize
S_person: contains a unique identifier for each person along with
demographic data.
S_ObsPeriod: contains the span of time for which data is captured in the
database for a given patient.
S_DrugExp: contains periods of drug usage for each person.
S_ConditionOCC: contains records of conditions and when they were
observed.
K-nearest Neighbor
After the extraction of the data, this information must be ran through
some sort of algorithm to distinguish what medications would be right
for a certain person. The algorithm that we chose was K Nearest
Neighbor and is the State-of- the- art algorithm used in recommender
systems. With this algorithm we are capable of finding similarities
between patients and predicting the best medication for them. Using
this algorithm also allowed use to create this type of vision with the data
we extracted: